Yutsukioka Un Job Application Helper Term Extractor

term_extractor

Purpose

This skill analyzes a job description to extract the five most important screening terms for ATS optimization and competency-based shortlisting. The output populates the TERM_EXTRACTOR section of inputs/application_context.md for use by downstream skills.

Important:

• Star ratings (★★★★★ etc.) are used here for prioritization.
• Do not paste star symbols into final application fields/documents unless explicitly requested.
• When other skills reference “★★/★★★ keywords”, they mean the underlying term (not the star symbols).

Apply the expert lens, collaboration rules, guardrails, quality loop protocol, and guiding principles defined in apex-guardrails.

Inputs

Required:

• JOB_DESCRIPTION_TEXT: the full job description including Responsibilities, Requirements/Qualifications, and Desirable sections.

Optional:

• USER_JOB_HISTORY_TEXT or candidate CV/LinkedIn bullets: to tailor resume examples. If missing, use placeholders like [country], [programme], [tool], [metric] and do not invent facts.
• Optimization preference: ATS-first / Human-first / Balanced (default: Balanced).

Success Metrics (optimize for all)

• M1: Evidence-Backed Relevance (9/10+): Each term is explicitly present in the JD or strongly implied by repeated duties. No guessing.
• M2: Coverage (8/10+): Across five terms, cover as many domains as the posting allows: core domain(s), technical skills/tools, UN frameworks/programmes/policies, high-impact deliverables, essential qualifications.
• M3: ATS Matching Strength (8/10+): Each term includes 6-12 synonyms with common variations and acronyms (if in the posting).
• M4: Non-Redundancy (9/10+): No near-duplicate terms unless the posting clearly treats them as separate requirements.
• M5: Format Compliance (10/10 required): Exactly five terms, each exactly four lines with required labels, separated by one blank line.

Star Rating Rubric (use ★ symbols)

• ★★★★★ = explicitly required / central to the role (title + required quals + repeated deliverables)
• ★★★★ = strongly emphasized; likely a screening keyword
• ★★★ = important but secondary
• ★★ = supportive / nice-to-have
• ★ = minor mention

Only use ★★★★★ when the posting clearly supports it.

Output format (STRIC)

For each of the five terms, output exactly four lines (no bullets, no numbering, no extra sections). Separate terms with one blank line.

• Line 1: <term> <stars>
• Line 2: Synonyms: syn1; syn2; syn3; ...
• Line 3: You should add this term because: <brief reason tied to JD>
• Line 4: Example for your resume: "one resume-style sentence"

No extra text before or after the five terms.

Steps

1. Extract (broad net): Identify 10-15 candidate terms/skill areas directly from the posting (title, duties, requirements). Keep them specific; avoid vague soft skills unless emphasized.
2. Score for screening likelihood (internal only; do not print): Assess each term for "Required/Essential/Must" language, repetition across sections, centrality to title + top duties, typical shortlist keywords, and qualification gatekeeping.
3. Select the final five: Choose the best 5 ensuring coverage (M2) and no redundancy (M4). If the posting includes a required degree/language/certification, ensure one term captures it.
4. Build synonyms: For each term, generate 6-12 semicolon-separated synonyms/variants, including acronyms and full names only if supported by the posting.
5. Justify with JD evidence: For each term, write a brief justification tied to where it appears (Responsibilities / Requirements / Desirable). May include a short JD phrase (max 8 words).
6. Write a resume-style example: Create one bullet-like sentence in quotes with action + scope + outcome. Use metrics if candidate info provides them; otherwise use placeholders.
7. Self-evaluation (internal only; do not print): Apply the recursive self-evaluation loop from apex-guardrails, adding domain-specific checks: verify M1-M5 thresholds, check for missed gatekeepers (degree/language/tools/frameworks), and eliminate redundancy or overly generic terms. Output only once all metrics pass.

Related skill (optional)

If you want a broader phrase bank (20–40 phrases) to support ATS insertion mapping beyond five terms, use: apex-jd-keyword-bank and paste its output into the JD_KEYWORD_BANK section of the context pack.

Converted and distributed by TomeVault — claim your Tome and manage your conversions.